Importing Libraries¶
In [ ]:
import os, shutil, pathlib
from tensorflow import keras
from tensorflow.keras import layers
import pathlib
from tensorflow.keras.utils import image_dataset_from_directory
import matplotlib.pyplot as plt
import seaborn as sns
import numpy as np
from PIL import Image
from pathlib import Path
from sklearn.metrics import confusion_matrix, classification_report, precision_recall_curve, auc, ConfusionMatrixDisplay
batch_size: The number of samples per gradient update.
img_height and img_width: The dimensions to which the images will be resized.
image_dataset_from_directory: This function loads images from the directory structure and automatically labels them based on the subdirectories.
Loading the data¶
In [ ]:
# This should point to the small dataset of the Kaggle Dogs vs Cats competition that was created in a previous notebook
data_folder = pathlib.Path("D:\\CSCN8010-LABS\\kaggle_dogs_vs_cats_small")
In [ ]:
train_dataset = image_dataset_from_directory(
data_folder / "train",
image_size=(180, 180),
batch_size=32)
validation_dataset = image_dataset_from_directory(
data_folder / "validation",
image_size=(180, 180),
batch_size=32)
test_dataset = image_dataset_from_directory(
data_folder / "test",
image_size=(180, 180),
batch_size=32)
Found 2000 files belonging to 2 classes. Found 1000 files belonging to 2 classes. Found 2000 files belonging to 2 classes.
EDA: Explore the data¶
In [ ]:
import matplotlib.pyplot as plt
import tensorflow as tf
from tensorflow.keras.utils import image_dataset_from_directory
import numpy as np
from pathlib import Path
# Load datasets
data_folder = Path("D:\\CSCN8010-LABS\\kaggle_dogs_vs_cats_small")
train_dataset = image_dataset_from_directory(data_folder / "train", image_size=(180, 180), batch_size=32)
validation_dataset = image_dataset_from_directory(data_folder / "validation", image_size=(180, 180), batch_size=32)
test_dataset = image_dataset_from_directory(data_folder / "test", image_size=(180, 180), batch_size=32)
# 1. Explore class distribution
def plot_class_distribution(dataset, title):
class_counts = {}
for _, labels in dataset:
unique, counts = np.unique(labels.numpy(), return_counts=True)
for label, count in zip(unique, counts):
class_counts[label] = class_counts.get(label, 0) + count
plt.bar(class_counts.keys(), class_counts.values(), tick_label=["Cats", "Dogs"])
plt.title(title)
plt.xlabel("Class")
plt.ylabel("Count")
plt.show()
plot_class_distribution(train_dataset, "Training Data Distribution")
plot_class_distribution(validation_dataset, "Validation Data Distribution")
plot_class_distribution(test_dataset, "Test Data Distribution")
# 2. Display a few sample images
def display_images(dataset, title):
plt.figure(figsize=(10, 10))
for images, labels in dataset.take(1):
for i in range(9):
ax = plt.subplot(3, 3, i + 1)
plt.imshow(images[i].numpy().astype("uint8"))
plt.title("Dog" if labels[i] == 1 else "Cat")
plt.axis("off")
plt.suptitle(title)
plt.show()
display_images(train_dataset, "Sample Training Images")
# 3. Statistics on Pixel Intensity
def compute_statistics(dataset, dataset_name):
means, stds = [], []
for images, _ in dataset:
means.append(np.mean(images.numpy() / 255.0, axis=(0, 1, 2)))
stds.append(np.std(images.numpy() / 255.0, axis=(0, 1, 2)))
print(f"{dataset_name} Mean Pixel Intensity: {np.mean(means, axis=0)}")
print(f"{dataset_name} Std. Dev. of Pixel Intensity: {np.mean(stds, axis=0)}")
compute_statistics(train_dataset, "Training Data")
compute_statistics(validation_dataset, "Validation Data")
compute_statistics(test_dataset, "Test Data")
Found 2000 files belonging to 2 classes. Found 1000 files belonging to 2 classes. Found 2000 files belonging to 2 classes.
Training Data Mean Pixel Intensity: [0.48080638 0.4516345 0.409062 ] Training Data Std. Dev. of Pixel Intensity: [0.25904176 0.25180757 0.25474995] Validation Data Mean Pixel Intensity: [0.4875215 0.45641398 0.4116987 ] Validation Data Std. Dev. of Pixel Intensity: [0.25802138 0.25197008 0.2550806 ] Test Data Mean Pixel Intensity: [0.48553178 0.45734078 0.4133416 ] Test Data Std. Dev. of Pixel Intensity: [0.2609459 0.2542641 0.2561265]
DEFINING THE MODEL CNN¶
In [ ]:
inputs = keras.Input(shape=(180, 180, 3))
x = layers.Rescaling(1./255)(inputs)
x = layers.Conv2D(filters=32, kernel_size=3, activation="relu")(x)
x = layers.MaxPooling2D(pool_size=2)(x)
x = layers.Conv2D(filters=64, kernel_size=3, activation="relu")(x)
x = layers.MaxPooling2D(pool_size=2)(x)
x = layers.Conv2D(filters=128, kernel_size=3, activation="relu")(x)
x = layers.MaxPooling2D(pool_size=2)(x)
x = layers.Conv2D(filters=256, kernel_size=3, activation="relu")(x)
x = layers.MaxPooling2D(pool_size=2)(x)
x = layers.Conv2D(filters=256, kernel_size=3, activation="relu")(x)
x = layers.Flatten()(x)
outputs = layers.Dense(1, activation="sigmoid")(x)
model = keras.Model(inputs=inputs, outputs=outputs)
In [ ]:
model.summary()
Model: "model"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
input_1 (InputLayer) [(None, 180, 180, 3)] 0
rescaling (Rescaling) (None, 180, 180, 3) 0
conv2d (Conv2D) (None, 178, 178, 32) 896
max_pooling2d (MaxPooling2D (None, 89, 89, 32) 0
)
conv2d_1 (Conv2D) (None, 87, 87, 64) 18496
max_pooling2d_1 (MaxPooling (None, 43, 43, 64) 0
2D)
conv2d_2 (Conv2D) (None, 41, 41, 128) 73856
max_pooling2d_2 (MaxPooling (None, 20, 20, 128) 0
2D)
conv2d_3 (Conv2D) (None, 18, 18, 256) 295168
max_pooling2d_3 (MaxPooling (None, 9, 9, 256) 0
2D)
conv2d_4 (Conv2D) (None, 7, 7, 256) 590080
flatten (Flatten) (None, 12544) 0
dense (Dense) (None, 1) 12545
=================================================================
Total params: 991,041
Trainable params: 991,041
Non-trainable params: 0
_________________________________________________________________
COMPILATION OF THE MODEL¶
In [ ]:
model.compile(loss="binary_crossentropy",
optimizer="rmsprop",
metrics=["accuracy"])
CALL BACKS FOR THE COMPILATION OF THE SAME BEST MODEL¶
In [ ]:
callbacks = [
keras.callbacks.ModelCheckpoint(
filepath="./models/convnet_from_scratch.keras",
save_best_only=True,
monitor="val_loss")
]
history = model.fit(
train_dataset,
epochs=30,
validation_data=validation_dataset,
callbacks=callbacks)
Epoch 1/30 63/63 [==============================] - 33s 507ms/step - loss: 0.8017 - accuracy: 0.5080 - val_loss: 0.6923 - val_accuracy: 0.5510 Epoch 2/30 63/63 [==============================] - 31s 488ms/step - loss: 0.6917 - accuracy: 0.5325 - val_loss: 0.6915 - val_accuracy: 0.5060 Epoch 3/30 63/63 [==============================] - 30s 475ms/step - loss: 0.6917 - accuracy: 0.5520 - val_loss: 0.6741 - val_accuracy: 0.6060 Epoch 4/30 63/63 [==============================] - 30s 471ms/step - loss: 0.6714 - accuracy: 0.6035 - val_loss: 0.6704 - val_accuracy: 0.5770 Epoch 5/30 63/63 [==============================] - 30s 472ms/step - loss: 0.6418 - accuracy: 0.6450 - val_loss: 0.6623 - val_accuracy: 0.6100 Epoch 6/30 63/63 [==============================] - 30s 475ms/step - loss: 0.6224 - accuracy: 0.6680 - val_loss: 0.6079 - val_accuracy: 0.6640 Epoch 7/30 63/63 [==============================] - 30s 469ms/step - loss: 0.5746 - accuracy: 0.7110 - val_loss: 0.6124 - val_accuracy: 0.6250 Epoch 8/30 63/63 [==============================] - 30s 471ms/step - loss: 0.5402 - accuracy: 0.7225 - val_loss: 0.5877 - val_accuracy: 0.6860 Epoch 9/30 63/63 [==============================] - 31s 495ms/step - loss: 0.5075 - accuracy: 0.7555 - val_loss: 0.6769 - val_accuracy: 0.6860 Epoch 10/30 63/63 [==============================] - 30s 480ms/step - loss: 0.4787 - accuracy: 0.7690 - val_loss: 0.5739 - val_accuracy: 0.7040 Epoch 11/30 63/63 [==============================] - 31s 486ms/step - loss: 0.4451 - accuracy: 0.7905 - val_loss: 0.6428 - val_accuracy: 0.7060 Epoch 12/30 63/63 [==============================] - 32s 501ms/step - loss: 0.3854 - accuracy: 0.8330 - val_loss: 0.6275 - val_accuracy: 0.7270 Epoch 13/30 63/63 [==============================] - 31s 487ms/step - loss: 0.3305 - accuracy: 0.8605 - val_loss: 0.8567 - val_accuracy: 0.6940 Epoch 14/30 63/63 [==============================] - 125s 2s/step - loss: 0.2745 - accuracy: 0.8795 - val_loss: 0.7007 - val_accuracy: 0.7220 Epoch 15/30 63/63 [==============================] - 149s 2s/step - loss: 0.2163 - accuracy: 0.9055 - val_loss: 0.8484 - val_accuracy: 0.7210 Epoch 16/30 63/63 [==============================] - 59s 946ms/step - loss: 0.1794 - accuracy: 0.9255 - val_loss: 0.8561 - val_accuracy: 0.6940 Epoch 17/30 63/63 [==============================] - 161s 3s/step - loss: 0.1466 - accuracy: 0.9470 - val_loss: 0.9408 - val_accuracy: 0.6930 Epoch 18/30 63/63 [==============================] - 163s 3s/step - loss: 0.1004 - accuracy: 0.9640 - val_loss: 1.1816 - val_accuracy: 0.7310 Epoch 19/30 63/63 [==============================] - 104s 2s/step - loss: 0.0974 - accuracy: 0.9620 - val_loss: 3.3659 - val_accuracy: 0.6130 Epoch 20/30 63/63 [==============================] - 165s 3s/step - loss: 0.1153 - accuracy: 0.9685 - val_loss: 1.4138 - val_accuracy: 0.7320 Epoch 21/30 63/63 [==============================] - 165s 3s/step - loss: 0.0643 - accuracy: 0.9800 - val_loss: 1.2629 - val_accuracy: 0.7210 Epoch 22/30 63/63 [==============================] - 103s 2s/step - loss: 0.0786 - accuracy: 0.9720 - val_loss: 1.3338 - val_accuracy: 0.7400 Epoch 23/30 63/63 [==============================] - 31s 495ms/step - loss: 0.0532 - accuracy: 0.9865 - val_loss: 1.5098 - val_accuracy: 0.7200 Epoch 24/30 63/63 [==============================] - 30s 480ms/step - loss: 0.0852 - accuracy: 0.9760 - val_loss: 1.9384 - val_accuracy: 0.6770 Epoch 25/30 63/63 [==============================] - 30s 472ms/step - loss: 0.0373 - accuracy: 0.9865 - val_loss: 1.8386 - val_accuracy: 0.7210 Epoch 26/30 63/63 [==============================] - 30s 481ms/step - loss: 0.0438 - accuracy: 0.9845 - val_loss: 2.2669 - val_accuracy: 0.6830 Epoch 27/30 63/63 [==============================] - 30s 473ms/step - loss: 0.0502 - accuracy: 0.9835 - val_loss: 2.2294 - val_accuracy: 0.7260 Epoch 28/30 63/63 [==============================] - 30s 471ms/step - loss: 0.0614 - accuracy: 0.9780 - val_loss: 2.2455 - val_accuracy: 0.7170 Epoch 29/30 63/63 [==============================] - 30s 476ms/step - loss: 0.0338 - accuracy: 0.9895 - val_loss: 2.0187 - val_accuracy: 0.7260 Epoch 30/30 63/63 [==============================] - 30s 480ms/step - loss: 0.0444 - accuracy: 0.9890 - val_loss: 2.0410 - val_accuracy: 0.7240
EVALUATING THE CNN MODEL¶
Displaying the curves of loss of acuracy during the training¶
In [ ]:
import matplotlib.pyplot as plt
# Train the model and store the history
history = model.fit(
train_dataset,
validation_data=validation_dataset,
epochs=10 # Number of epochs to train
)
# Extract accuracy and loss from the history object
accuracy = history.history["accuracy"]
val_accuracy = history.history["val_accuracy"]
loss = history.history["loss"]
val_loss = history.history["val_loss"]
# Create the epochs range
epochs = range(1, len(accuracy) + 1)
# Plot training and validation accuracy
plt.plot(epochs, accuracy, "bo", label="Training accuracy")
plt.plot(epochs, val_accuracy, "b", label="Validation accuracy")
plt.title("Training and Validation Accuracy")
plt.legend()
plt.figure()
# Plot training and validation loss
plt.plot(epochs, loss, "bo", label="Training loss")
plt.plot(epochs, val_loss, "b", label="Validation loss")
plt.title("Training and Validation Loss")
plt.legend()
plt.show()
Epoch 1/10 63/63 [==============================] - 32s 502ms/step - loss: 0.0453 - accuracy: 0.9830 - val_loss: 2.2628 - val_accuracy: 0.7230 Epoch 2/10 63/63 [==============================] - 31s 497ms/step - loss: 0.0424 - accuracy: 0.9850 - val_loss: 2.5342 - val_accuracy: 0.7150 Epoch 3/10 63/63 [==============================] - 31s 495ms/step - loss: 0.0601 - accuracy: 0.9870 - val_loss: 1.7578 - val_accuracy: 0.7390 Epoch 4/10 63/63 [==============================] - 33s 523ms/step - loss: 0.0688 - accuracy: 0.9835 - val_loss: 2.2058 - val_accuracy: 0.6960 Epoch 5/10 63/63 [==============================] - 32s 503ms/step - loss: 0.0418 - accuracy: 0.9870 - val_loss: 2.3124 - val_accuracy: 0.7250 Epoch 6/10 63/63 [==============================] - 31s 495ms/step - loss: 0.0291 - accuracy: 0.9880 - val_loss: 2.4132 - val_accuracy: 0.7170 Epoch 7/10 63/63 [==============================] - 31s 491ms/step - loss: 0.0318 - accuracy: 0.9890 - val_loss: 2.5252 - val_accuracy: 0.7100 Epoch 8/10 63/63 [==============================] - 32s 509ms/step - loss: 0.0572 - accuracy: 0.9805 - val_loss: 2.7229 - val_accuracy: 0.7180 Epoch 9/10 63/63 [==============================] - 33s 516ms/step - loss: 0.0335 - accuracy: 0.9915 - val_loss: 3.3099 - val_accuracy: 0.7050 Epoch 10/10 63/63 [==============================] - 32s 509ms/step - loss: 0.0443 - accuracy: 0.9895 - val_loss: 2.7990 - val_accuracy: 0.7070
TEST ACUURACY¶
In [ ]:
test_model = keras.models.load_model("./models/convnet_from_scratch.keras")
test_loss, test_acc = test_model.evaluate(test_dataset)
print(f"Test accuracy: {test_acc:.3f}")
63/63 [==============================] - 7s 114ms/step - loss: 0.5888 - accuracy: 0.7015 Test accuracy: 0.701
In [ ]:
test_model = keras.models.load_model("./models/convnet_from_scratch.keras")
test_loss, test_acc = test_model.evaluate(test_dataset)
print(f"Test accuracy: {test_acc:.3f}")
63/63 [==============================] - 8s 118ms/step - loss: 0.5888 - accuracy: 0.7015 Test accuracy: 0.701
Confusion - Matrix¶
In [ ]:
import tensorflow as tf
from sklearn.metrics import confusion_matrix, ConfusionMatrixDisplay
import numpy as np
import matplotlib.pyplot as plt
# Assume the model is already trained
# model = ... (Your trained model)
# 1. Get actual labels and predictions
def get_actual_and_predicted_labels(model, dataset):
y_true = []
y_pred = []
for images, labels in dataset:
predictions = model.predict(images) # Get predicted probabilities
predicted_labels = np.round(predictions).flatten() # Convert to binary labels
y_true.extend(labels.numpy())
y_pred.extend(predicted_labels)
return np.array(y_true), np.array(y_pred)
# Get the labels and predictions
y_true, y_pred = get_actual_and_predicted_labels(model, test_dataset)
# 2. Generate confusion matrix
conf_matrix = confusion_matrix(y_true, y_pred)
# 3. Display the confusion matrix
disp = ConfusionMatrixDisplay(confusion_matrix=conf_matrix, display_labels=["Cat", "Dog"])
disp.plot(cmap=plt.cm.Blues)
plt.title("Confusion Matrix")
plt.show()
# 4. Print metrics from confusion matrix
TP = conf_matrix[1, 1] # True Positives
TN = conf_matrix[0, 0] # True Negatives
FP = conf_matrix[0, 1] # False Positives
FN = conf_matrix[1, 0] # False Negatives
print(f"True Positives (Dogs correctly classified): {TP}")
print(f"True Negatives (Cats correctly classified): {TN}")
print(f"False Positives (Cats classified as Dogs): {FP}")
print(f"False Negatives (Dogs classified as Cats): {FN}")
1/1 [==============================] - 0s 218ms/step 1/1 [==============================] - 0s 125ms/step 1/1 [==============================] - 0s 122ms/step 1/1 [==============================] - 0s 107ms/step 1/1 [==============================] - 0s 120ms/step 1/1 [==============================] - 0s 124ms/step 1/1 [==============================] - 0s 132ms/step 1/1 [==============================] - 0s 121ms/step 1/1 [==============================] - 0s 115ms/step 1/1 [==============================] - 0s 122ms/step 1/1 [==============================] - 0s 136ms/step 1/1 [==============================] - 0s 128ms/step 1/1 [==============================] - 0s 113ms/step 1/1 [==============================] - 0s 131ms/step 1/1 [==============================] - 0s 127ms/step 1/1 [==============================] - 0s 116ms/step 1/1 [==============================] - 0s 115ms/step 1/1 [==============================] - 0s 126ms/step 1/1 [==============================] - 0s 131ms/step 1/1 [==============================] - 0s 127ms/step 1/1 [==============================] - 0s 117ms/step 1/1 [==============================] - 0s 110ms/step 1/1 [==============================] - 0s 110ms/step 1/1 [==============================] - 0s 100ms/step 1/1 [==============================] - 0s 110ms/step 1/1 [==============================] - 0s 111ms/step 1/1 [==============================] - 0s 117ms/step 1/1 [==============================] - 0s 118ms/step 1/1 [==============================] - 0s 118ms/step 1/1 [==============================] - 0s 114ms/step 1/1 [==============================] - 0s 124ms/step 1/1 [==============================] - 0s 120ms/step 1/1 [==============================] - 0s 114ms/step 1/1 [==============================] - 0s 113ms/step 1/1 [==============================] - 0s 104ms/step 1/1 [==============================] - 0s 114ms/step 1/1 [==============================] - 0s 111ms/step 1/1 [==============================] - 0s 119ms/step 1/1 [==============================] - 0s 113ms/step 1/1 [==============================] - 0s 108ms/step 1/1 [==============================] - 0s 103ms/step 1/1 [==============================] - 0s 110ms/step 1/1 [==============================] - 0s 108ms/step 1/1 [==============================] - 0s 100ms/step 1/1 [==============================] - 0s 113ms/step 1/1 [==============================] - 0s 114ms/step 1/1 [==============================] - 0s 115ms/step 1/1 [==============================] - 0s 120ms/step 1/1 [==============================] - 0s 106ms/step 1/1 [==============================] - 0s 107ms/step 1/1 [==============================] - 0s 104ms/step 1/1 [==============================] - 0s 107ms/step 1/1 [==============================] - 0s 129ms/step 1/1 [==============================] - 0s 102ms/step 1/1 [==============================] - 0s 104ms/step 1/1 [==============================] - 0s 107ms/step 1/1 [==============================] - 0s 120ms/step 1/1 [==============================] - 0s 114ms/step 1/1 [==============================] - 0s 107ms/step 1/1 [==============================] - 0s 108ms/step 1/1 [==============================] - 0s 112ms/step 1/1 [==============================] - 0s 104ms/step 1/1 [==============================] - 0s 109ms/step
True Positives (Dogs correctly classified): 746 True Negatives (Cats correctly classified): 688 False Positives (Cats classified as Dogs): 312 False Negatives (Dogs classified as Cats): 254
Classification Report¶
In [ ]:
from sklearn.metrics import classification_report
def calculate_metrics(y_true, y_pred):
print("Classification Report:")
print(classification_report(y_true, y_pred, target_names=["Cat", "Dog"]))
calculate_metrics(y_true, y_pred)
Classification Report:
precision recall f1-score support
Cat 0.73 0.69 0.71 1000
Dog 0.71 0.75 0.72 1000
accuracy 0.72 2000
macro avg 0.72 0.72 0.72 2000
weighted avg 0.72 0.72 0.72 2000
Precision Recall Curve¶
In [ ]:
from sklearn.metrics import precision_recall_curve, auc
def precision_recall_curve_analysis(model, dataset):
y_true, y_prob = [], []
for images, labels in dataset:
probs = model.predict(images).flatten()
y_true.extend(labels.numpy())
y_prob.extend(probs)
y_true = np.array(y_true)
y_prob = np.array(y_prob)
precision, recall, _ = precision_recall_curve(y_true, y_prob)
pr_auc = auc(recall, precision)
# Plot Precision-Recall Curve
plt.plot(recall, precision, label=f"PR Curve (AUC={pr_auc:.2f})")
plt.xlabel("Recall")
plt.ylabel("Precision")
plt.title("Precision-Recall Curve")
plt.legend()
plt.show()
precision_recall_curve_analysis(model, test_dataset)
1/1 [==============================] - 0s 111ms/step 1/1 [==============================] - 0s 97ms/step 1/1 [==============================] - 0s 85ms/step 1/1 [==============================] - 0s 90ms/step 1/1 [==============================] - 0s 91ms/step 1/1 [==============================] - 0s 96ms/step 1/1 [==============================] - 0s 89ms/step 1/1 [==============================] - 0s 94ms/step 1/1 [==============================] - 0s 90ms/step 1/1 [==============================] - 0s 95ms/step 1/1 [==============================] - 0s 97ms/step 1/1 [==============================] - 0s 83ms/step 1/1 [==============================] - 0s 107ms/step 1/1 [==============================] - 0s 105ms/step 1/1 [==============================] - 0s 102ms/step 1/1 [==============================] - 0s 115ms/step 1/1 [==============================] - 0s 96ms/step 1/1 [==============================] - 0s 96ms/step 1/1 [==============================] - 0s 94ms/step 1/1 [==============================] - 0s 91ms/step 1/1 [==============================] - 0s 105ms/step 1/1 [==============================] - 0s 104ms/step 1/1 [==============================] - 0s 108ms/step 1/1 [==============================] - 0s 97ms/step 1/1 [==============================] - 0s 102ms/step 1/1 [==============================] - 0s 100ms/step 1/1 [==============================] - 0s 91ms/step 1/1 [==============================] - 0s 97ms/step 1/1 [==============================] - 0s 102ms/step 1/1 [==============================] - 0s 97ms/step 1/1 [==============================] - 0s 96ms/step 1/1 [==============================] - 0s 106ms/step 1/1 [==============================] - 0s 93ms/step 1/1 [==============================] - 0s 98ms/step 1/1 [==============================] - 0s 99ms/step 1/1 [==============================] - 0s 95ms/step 1/1 [==============================] - 0s 91ms/step 1/1 [==============================] - 0s 78ms/step 1/1 [==============================] - 0s 79ms/step 1/1 [==============================] - 0s 95ms/step 1/1 [==============================] - 0s 102ms/step 1/1 [==============================] - 0s 87ms/step 1/1 [==============================] - 0s 77ms/step 1/1 [==============================] - 0s 89ms/step 1/1 [==============================] - 0s 72ms/step 1/1 [==============================] - 0s 78ms/step 1/1 [==============================] - 0s 81ms/step 1/1 [==============================] - 0s 103ms/step 1/1 [==============================] - 0s 93ms/step 1/1 [==============================] - 0s 102ms/step 1/1 [==============================] - 0s 84ms/step 1/1 [==============================] - 0s 85ms/step 1/1 [==============================] - 0s 82ms/step 1/1 [==============================] - 0s 74ms/step 1/1 [==============================] - 0s 86ms/step 1/1 [==============================] - 0s 99ms/step 1/1 [==============================] - 0s 91ms/step 1/1 [==============================] - 0s 88ms/step 1/1 [==============================] - 0s 91ms/step 1/1 [==============================] - 0s 91ms/step 1/1 [==============================] - 0s 96ms/step 1/1 [==============================] - 0s 101ms/step 1/1 [==============================] - 0s 54ms/step
Explore Misclassified Example¶
In [ ]:
def explore_misclassified_examples(y_true, y_pred, dataset):
misclassified_indices = np.where(y_true != y_pred)[0]
print(f"Number of misclassified examples: {len(misclassified_indices)}")
# Display a few misclassified images
for i in misclassified_indices[:5]: # Show first 5 examples
for images, labels in dataset.take(i+1):
misclassified_image = images[0].numpy().astype("uint8")
true_label = labels[0].numpy()
predicted_label = y_pred[i]
plt.imshow(misclassified_image)
plt.title(f"True: {'Dog' if true_label == 1 else 'Cat'}, Predicted: {'Dog' if predicted_label == 1 else 'Cat'}")
plt.axis("off")
plt.show()
explore_misclassified_examples(y_true, y_pred, test_dataset)
Number of misclassified examples: 566
CONCLUSION¶
From the above analyses,
Accuracy and Confusion Matrix: Overall performance assessment and particular misclassifications-for example, more false positives or false negatives.
Precision, Recall, and F1-score: Summarize the model's performance in identifying the positive class, Dogs.
Precision-Recall Curve: This is done to evaluate trade-offs at different thresholds that yield the best balance between precision and recall.
Error Analysis Identify whether the model struggles with ambiguous images or features of specific type, for instance, small dogs or blurry cats.